Security audit

个性化新闻助手免费版

Security checks across malware telemetry and agentic risk

Overview

This is a coherent personalized news briefing skill, but users should understand it stores local preference/history files and web searches may reveal their interests to search providers.

Install only if you are comfortable with the agent keeping a local ~/news profile and history of your news interests. Treat the privacy claim as local-storage-focused, not strictly offline: generating briefings uses web search, so search terms based on your interests may be sent to external services. Review or delete the ~/news files if you do not want retained preferences or reading history.

SkillSpector

Vulnerability Patterns

Trigger AbuseOverly Broad Trigger, Shadow Command Trigger, Keyword Baiting Trigger
MCP Tool PoisoningHidden Instructions, Unicode Deception, Parameter Description Injection
Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands
Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
Privilege EscalationExcessive Permissions, Sudo/Root Execution, Credential Access

Findings (5)

Description-Behavior Mismatch

Medium

Confidence: 90% confidence
Finding: The skill promises that user data remains local and is not uploaded, yet its core functionality depends on internet-based news retrieval through external search services. Even if the implementation does not intentionally exfiltrate data, user interests, prompts, and query context may still be sent to third-party providers, creating a privacy and transparency gap.

Intent-Code Divergence

Medium

Confidence: 81% confidence
Finding: The document says the free version does not support custom news-source integration, but it also provides editable source configuration with external URLs. This mismatch can mislead users or downstream agents into treating the source list as trusted and fixed when it can in fact be modified, enabling unreviewed or malicious sources to influence outputs.

Vague Triggers

High

Confidence: 92% confidence
Finding: The trigger conditions are inconsistent with the skill's stated purpose and include broad content-generation scenarios such as marketing copy, title optimization, and general writing. Overbroad activation rules can cause the agent to invoke this skill outside its intended domain, leading to inappropriate tool use, unintended file writes, privacy leakage, or unsafe delegation based on irrelevant requests.

Vague Triggers

Medium

Confidence: 84% confidence
Finding: The skill declares coverage using a long list of broad keywords without meaningful scope constraints. This increases the chance that an orchestrator or agent router will match and run the skill on loosely related requests, causing unexpected behavior, privacy-impacting local storage actions, or retrieval against external services when not actually needed.

Missing User Warnings

Medium

Confidence: 87% confidence
Finding: The skill instructs the agent to create and update local profile and history files containing user preferences and reading behavior, but it does not require explicit notice, consent, or confirmation before modifying persistent data. This can lead to silent retention of sensitive behavioral data and make users unaware of what is stored or how it may later influence outputs.

VirusTotal

64/64 vendors flagged this skill as clean.

View on VirusTotal

Static analysis

No suspicious patterns detected.